In [3]: import warnings

warnings.filterwarnings('ignore')

In [4]: import pandas as pd

import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline

In [5]: #uploadind data file

loan = pd.read_csv(r"C:\Users\dell\Downloads\loan (1).zip")

In [6]: loan.head(10)

Out[6]: id member_id loan_amnt funded_amnt funded_amnt_inv term int_rate installment grade sub_gr

36
0 1077501 1296599 5000 5000 4975.0 10.65% 162.87 B
months

60
1 1077430 1314167 2500 2500 2500.0 15.27% 59.83 C
months

36
2 1077175 1313524 2400 2400 2400.0 15.96% 84.33 C
months

36
3 1076863 1277178 10000 10000 10000.0 13.49% 339.31 C
months

60
4 1075358 1311748 3000 3000 3000.0 12.69% 67.79 B
months

36
5 1075269 1311441 5000 5000 5000.0 7.90% 156.46 A
months

60
6 1069639 1304742 7000 7000 7000.0 15.96% 170.08 C
months

36
7 1072053 1288686 3000 3000 3000.0 18.64% 109.43 E
months

60
8 1071795 1306957 5600 5600 5600.0 21.28% 152.39 F
months

60
9 1071570 1306721 5375 5375 5350.0 12.69% 121.45 B
months

10 rows × 111 columns

In [7]: loan.shape

(39717, 111)
Out[7]:

In [8]: loan.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 39717 entries, 0 to 39716
Columns: 111 entries, id to total_il_high_credit_limit
dtypes: float64(74), int64(13), object(24)
memory usage: 33.6+ MB

In [9]: loan.isnull().any()

Loading [MathJax]/extensions/Safe.js
 id False
Out[9]:
member_id False
loan_amnt False
funded_amnt False
funded_amnt_inv False
...
tax_liens True
tot_hi_cred_lim True
total_bal_ex_mort True
total_bc_limit True
total_il_high_credit_limit True
Length: 111, dtype: bool

In [10]: loan.isnull().any().sum()

68
Out[10]:

In [11]: loan.describe()

Out[11]: id member_id loan_amnt funded_amnt funded_amnt_inv installment annual_inc

count 3.971700e+04 3.971700e+04 39717.000000 39717.000000 39717.000000 39717.000000 3.971700e+04

mean 6.831319e+05 8.504636e+05 11219.443815 10947.713196 10397.448868 324.561922 6.896893e+04

std 2.106941e+05 2.656783e+05 7456.670694 7187.238670 7128.450439 208.874874 6.379377e+04

min 5.473400e+04 7.069900e+04 500.000000 500.000000 0.000000 15.690000 4.000000e+03

25% 5.162210e+05 6.667800e+05 5500.000000 5400.000000 5000.000000 167.020000 4.040400e+04

50% 6.656650e+05 8.508120e+05 10000.000000 9600.000000 8975.000000 280.220000 5.900000e+04

75% 8.377550e+05 1.047339e+06 15000.000000 15000.000000 14400.000000 430.780000 8.230000e+04

max 1.077501e+06 1.314167e+06 35000.000000 35000.000000 35000.000000 1305.190000 6.000000e+06

8 rows × 87 columns

In [12]: loan.describe(include = 'all')

Out[12]: id member_id loan_amnt funded_amnt funded_amnt_inv term int_rate installme

count 3.971700e+04 3.971700e+04 39717.000000 39717.000000 39717.000000 39717 39717 39717.0000

unique NaN NaN NaN NaN NaN 2 371 Na

36
top NaN NaN NaN NaN NaN 10.99% Na
months

freq NaN NaN NaN NaN NaN 29096 956 Na

mean 6.831319e+05 8.504636e+05 11219.443815 10947.713196 10397.448868 NaN NaN 324.5619

std 2.106941e+05 2.656783e+05 7456.670694 7187.238670 7128.450439 NaN NaN 208.8748

min 5.473400e+04 7.069900e+04 500.000000 500.000000 0.000000 NaN NaN 15.6900

25% 5.162210e+05 6.667800e+05 5500.000000 5400.000000 5000.000000 NaN NaN 167.0200

50% 6.656650e+05 8.508120e+05 10000.000000 9600.000000 8975.000000 NaN NaN 280.2200

75% 8.377550e+05 1.047339e+06 15000.000000 15000.000000 14400.000000 NaN NaN 430.7800

max 1.077501e+06 1.314167e+06 35000.000000 35000.000000 35000.000000 NaN NaN 1305.1900

11 rows × 111 columns

Loading [MathJax]/extensions/Safe.js
 In [13]: columns = loan.columns

In [14]: percent_missing = round(100*loan.isnull().sum()/len(loan.index),2)

In [15]: missing_val_df = pd.DataFrame({'column_name': columns , 'percent_missing':percent_missin

In [16]: drop_missing = missing_val_df[missing_val_df.percent_missing == 100].column_name

In [17]: loan = loan.drop(drop_missing, axis = 1)

In [18]: loan

Out[18]: id member_id loan_amnt funded_amnt funded_amnt_inv term int_rate installment grade su

36
0 1077501 1296599 5000 5000 4975.0 10.65% 162.87 B
months

60
1 1077430 1314167 2500 2500 2500.0 15.27% 59.83 C
months

36
2 1077175 1313524 2400 2400 2400.0 15.96% 84.33 C
months

36
3 1076863 1277178 10000 10000 10000.0 13.49% 339.31 C
months

60
4 1075358 1311748 3000 3000 3000.0 12.69% 67.79 B
months

... ... ... ... ... ... ... ... ... ...

36
39712 92187 92174 2500 2500 1075.0 8.07% 78.42 A
months

36
39713 90665 90607 8500 8500 875.0 10.28% 275.38 C
months

36
39714 90395 90390 5000 5000 1325.0 8.07% 156.84 A
months

36
39715 90376 89243 5000 5000 650.0 7.43% 155.38 A
months

36
39716 87023 86999 7500 7500 800.0 13.75% 255.43 E
months

39717 rows × 57 columns

In [19]: loan.columns

Index(['id', 'member_id', 'loan_amnt', 'funded_amnt', 'funded_amnt_inv',

Out[19]:
'term', 'int_rate', 'installment', 'grade', 'sub_grade', 'emp_title',
'emp_length', 'home_ownership', 'annual_inc', 'verification_status',
'issue_d', 'loan_status', 'pymnt_plan', 'url', 'desc', 'purpose',
'title', 'zip_code', 'addr_state', 'dti', 'delinq_2yrs',
'earliest_cr_line', 'inq_last_6mths', 'mths_since_last_delinq',
'mths_since_last_record', 'open_acc', 'pub_rec', 'revol_bal',
'revol_util', 'total_acc', 'initial_list_status', 'out_prncp',
'out_prncp_inv', 'total_pymnt', 'total_pymnt_inv', 'total_rec_prncp',
'total_rec_int', 'total_rec_late_fee', 'recoveries',
'collection_recovery_fee', 'last_pymnt_d', 'last_pymnt_amnt',
'next_pymnt_d', 'last_credit_pull_d', 'collections_12_mths_ex_med',
'policy_code', 'application_type', 'acc_now_delinq',
'chargeoff_within_12_mths', 'delinq_amnt', 'pub_rec_bankruptcies',
'tax_liens'],
dtype='object')
Loading [MathJax]/extensions/Safe.js
 In [20]: #Personal Informtion of customers, and is not usefull for EDA purpose

loan.drop(['title', 'emp_title','url', 'zip_code', 'desc'], axis = 1, inplace = True)

In [21]: loan.drop(['last_credit_pull_d','last_pymnt_amnt','last_pymnt_d','collection_recovery_fe
'total_rec_late_fee', 'total_rec_int','total_rec_prncp', 'total_pymnt_inv', 'o

loan.shape

(39717, 41)
Out[21]:

In [22]: loan.columns

Index(['id', 'member_id', 'loan_amnt', 'funded_amnt', 'funded_amnt_inv',

Out[22]:
'term', 'int_rate', 'installment', 'grade', 'sub_grade', 'emp_length',
'home_ownership', 'annual_inc', 'verification_status', 'issue_d',
'loan_status', 'pymnt_plan', 'purpose', 'addr_state', 'dti',
'delinq_2yrs', 'earliest_cr_line', 'inq_last_6mths',
'mths_since_last_delinq', 'mths_since_last_record', 'open_acc',
'pub_rec', 'revol_bal', 'revol_util', 'total_acc',
'initial_list_status', 'total_pymnt', 'next_pymnt_d',
'collections_12_mths_ex_med', 'policy_code', 'application_type',
'acc_now_delinq', 'chargeoff_within_12_mths', 'delinq_amnt',
'pub_rec_bankruptcies', 'tax_liens'],
dtype='object')

In [23]: loan.drop(['installment','pymnt_plan',
'delinq_2yrs', 'earliest_cr_line', 'inq_last_6mths',
'mths_since_last_delinq', 'mths_since_last_record', 'open_acc',
'pub_rec', 'revol_bal', 'revol_util', 'total_acc',
'initial_list_status', 'total_pymnt', 'next_pymnt_d',
'collections_12_mths_ex_med', 'policy_code', 'application_type',
'acc_now_delinq', 'chargeoff_within_12_mths', 'delinq_amnt',
'pub_rec_bankruptcies', 'tax_liens'], axis=1, inplace = True)

In [24]: loan.shape

(39717, 18)
Out[24]:

In [25]: loan.columns

Index(['id', 'member_id', 'loan_amnt', 'funded_amnt', 'funded_amnt_inv',

Out[25]:
'term', 'int_rate', 'grade', 'sub_grade', 'emp_length',
'home_ownership', 'annual_inc', 'verification_status', 'issue_d',
'loan_status', 'purpose', 'addr_state', 'dti'],
dtype='object')

In [26]: #Check if the there is any repeated loan ID

loan['id'].duplicated().sum()

0
Out[26]:

In [27]: #Reoving id's, they are'nt useful for analysis.

loan.drop(['id', 'member_id'], axis = 1, inplace = True)

In [28]: loan.shape

(39717, 16)
Out[28]:

In [29]: loan.head()
Loading [MathJax]/extensions/Safe.js
 Out[29]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

36
0 5000 5000 4975.0 10.65% B B2 10+ years RENT
months

60
1 2500 2500 2500.0 15.27% C C4 < 1 year RENT
months

36
2 2400 2400 2400.0 15.96% C C5 10+ years RENT
months

36
3 10000 10000 10000.0 13.49% C C1 10+ years RENT
months

60
4 3000 3000 3000.0 12.69% B B5 1 year RENT
months

Since we will be analysing the scenario where the chances are more that a loan applicant will default.
Hence, we will ignore the loans that are currently running and therefore, we will use loans which are fully
paid and charged off for the analysis.

In [30]: loan = loan[loan.loan_status != 'Current']

In [31]: #Checking the null values

round(100*loan.isnull().sum()/len(loan.index),2)

loan_amnt 0.00
Out[31]:
funded_amnt 0.00
funded_amnt_inv 0.00
term 0.00
int_rate 0.00
grade 0.00
sub_grade 0.00
emp_length 2.68
home_ownership 0.00
annual_inc 0.00
verification_status 0.00
issue_d 0.00
loan_status 0.00
purpose 0.00
addr_state 0.00
dti 0.00
dtype: float64

In [32]: #Removing the null values from emp_length as the null percentage is less than 5%, hence i
loan = loan[~loan['emp_length'].isnull()]

checking the purpose of loan as this is the important factor to be used in analysis.

In [33]: round(loan.purpose.value_counts(normalize = True)*100, 2)

Loading [MathJax]/extensions/Safe.js
 debt_consolidation 47.08
Out[33]:
credit_card 13.05
other 9.89
home_improvement 7.42
major_purchase 5.54
small_business 4.55
car 3.86
wedding 2.43
medical 1.75
moving 1.47
house 0.94
vacation 0.93
educational 0.84
renewable_energy 0.25
Name: purpose, dtype: float64

In [34]: #Since, we do not know what the terms other stands for, so get ris of it.
loan.drop(loan[loan.purpose == 'other'].index, inplace = True)

In [35]: loan.purpose.value_counts()

debt_consolidation 17675
Out[35]:
credit_card 4899
home_improvement 2785
major_purchase 2080
small_business 1710
car 1448
wedding 913
medical 656
moving 552
house 354
vacation 348
educational 317
renewable_energy 94
Name: purpose, dtype: int64

In [36]: loan.term.value_counts()

36 months 25270
Out[36]:
60 months 8561
Name: term, dtype: int64

In [37]: loan.term = loan.term.apply(lambda x: int(x.split()[0]) if x.find('month')>0 else int(x

loan.head(5)

Out[37]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65% B B2 10+ years RENT

1 2500 2500 2500.0 60 15.27% C C4 < 1 year RENT

2 2400 2400 2400.0 36 15.96% C C5 10+ years RENT

5 5000 5000 5000.0 36 7.90% A A4 3 years RENT

6 7000 7000 7000.0 60 15.96% C C5 8 years RENT

In [38]: loan['int_rate'].value_counts()

Loading [MathJax]/extensions/Safe.js
 10.99% 800
Out[38]:
11.49% 685
13.49% 669
7.51% 660
7.88% 637
...
16.20% 1
18.72% 1
16.01% 1
16.96% 1
14.70% 1
Name: int_rate, Length: 367, dtype: int64

In [39]: loan['int_rate'] = loan['int_rate'].str.replace('%','')

loan['int_rate'] = loan['int_rate'].astype(float)

loan.dtypes

loan_amnt int64
Out[39]:
funded_amnt int64
funded_amnt_inv float64
term int64
int_rate float64
grade object
sub_grade object
emp_length object
home_ownership object
annual_inc float64
verification_status object
issue_d object
loan_status object
purpose object
addr_state object
dti float64
dtype: object

In [40]: loan.head(5)

Out[40]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65 B B2 10+ years RENT

1 2500 2500 2500.0 60 15.27 C C4 < 1 year RENT

2 2400 2400 2400.0 36 15.96 C C5 10+ years RENT

5 5000 5000 5000.0 36 7.90 A A4 3 years RENT

6 7000 7000 7000.0 60 15.96 C C5 8 years RENT

In [41]: loan.emp_length.value_counts()

10+ years 7664

Out[41]:
< 1 year 3994
2 years 3847
3 years 3636
4 years 3027
5 years 2895
1 year 2811
6 years 1987
7 years 1550
8 years 1308
9 years 1112
Name: emp_length, dtype: int64
Loading [MathJax]/extensions/Safe.js
 In [42]: #Cleaning the emp_length and converting it to int
#Should be treating 0 and <1 as 0 as they all fall under less than 1 year category
#Should be treating 10+ years as 10 which would mean 10 years and above
loan['emp_length'] = loan['emp_length'].str.replace('+', '')
loan['emp_length'] =loan['emp_length'].str.replace('< 1 ', '0')
loan['emp_length'] =loan['emp_length'].str.replace('years','')
loan['emp_length'] =loan['emp_length'].str.replace('year','')
loan['emp_length'] = loan['emp_length'].astype(int)

In [43]: loan.emp_length.value_counts()

10 7664
Out[43]:
0 3994
2 3847
3 3636
4 3027
5 2895
1 2811
6 1987
7 1550
8 1308
9 1112
Name: emp_length, dtype: int64

In [44]: loan.head(5)

Out[44]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65 B B2 10 RENT

1 2500 2500 2500.0 60 15.27 C C4 0 RENT

2 2400 2400 2400.0 36 15.96 C C5 10 RENT

5 5000 5000 5000.0 36 7.90 A A4 3 RENT

6 7000 7000 7000.0 60 15.96 C C5 8 RENT

In [45]: loan['annual_inc'].describe()

count 3.383100e+04
Out[45]:
mean 7.003334e+04
std 6.596019e+04
min 4.000000e+03
25% 4.200000e+04
50% 6.000000e+04
75% 8.400000e+04
max 6.000000e+06
Name: annual_inc, dtype: float64

In [46]: loan['annual_inc'].quantile([0.5, 0.7,0.9, 0.95, 0.99, 1])

0.50 60000.0
Out[46]:
0.70 77000.0
0.90 118000.0
0.95 143387.5
0.99 235560.0
1.00 6000000.0
Name: annual_inc, dtype: float64

In [47]: quartile = loan['annual_inc'].quantile(0.99)

loan = loan[loan['annual_inc'] < quartile]

loan['annual_inc'].describe()
Loading [MathJax]/extensions/Safe.js
 count 33492.000000
Out[47]:
mean 66490.324304
std 35165.906151
min 4000.000000
25% 42000.000000
50% 60000.000000
75% 82000.000000
max 235000.000000
Name: annual_inc, dtype: float64

In [48]: loan['issue_d'].value_counts()

Loading [MathJax]/extensions/Safe.js
 Dec-11 1853
Out[48]:
Nov-11 1791
Sep-11 1670
Oct-11 1661
Aug-11 1569
Jun-11 1479
Jul-11 1474
May-11 1419
Apr-11 1371
Mar-11 1304
Jan-11 1244
Feb-11 1154
Dec-10 1116
Oct-10 1004
Nov-10 1000
Jul-10 971
Sep-10 955
Aug-10 944
Jun-10 850
May-10 771
Apr-10 667
Mar-10 599
Feb-10 506
Nov-09 504
Jan-10 484
Dec-09 473
Oct-09 464
Sep-09 407
Aug-09 363
Jul-09 331
Jun-09 318
May-09 283
Mar-09 251
Apr-09 246
Feb-09 228
Jan-09 221
Dec-08 203
Mar-08 198
Nov-08 167
Feb-08 154
Jan-08 145
Apr-08 125
Oct-08 82
Dec-07 74
Jul-08 65
May-08 62
Aug-08 57
Jun-08 56
Oct-07 31
Nov-07 29
Aug-07 29
Jul-07 28
Sep-08 27
Sep-07 14
Jun-07 1
Name: issue_d, dtype: int64

In [49]: loan['issue_month'] = loan['issue_d'].apply(lambda x: str(x.split('-')[0])if x.find('-')

loan['issue_year'] = loan['issue_d'].apply(lambda x: str(x.split('-')[1])if x.find('-')

loan['issue_year'] = '20'+ loan['issue_year']

loan.head(5)
Loading [MathJax]/extensions/Safe.js
 Out[49]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65 B B2 10 RENT

1 2500 2500 2500.0 60 15.27 C C4 0 RENT

2 2400 2400 2400.0 36 15.96 C C5 10 RENT

5 5000 5000 5000.0 36 7.90 A A4 3 RENT

6 7000 7000 7000.0 60 15.96 C C5 8 RENT

In [50]: loan['loan_inc_ratio'] = round(100*(loan['loan_amnt']/loan['annual_inc']),2)

loan.head(2)

Out[50]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65 B B2 10 RENT

1 2500 2500 2500.0 60 15.27 C C4 0 RENT

In [51]: loan['loan_inc_ratio'].quantile([.25,.50,.75])

0.25 10.29
Out[51]:
0.50 16.67
0.75 25.38
Name: loan_inc_ratio, dtype: float64

In [52]: # Categorising loan_inc_ratio into categorised_loan_inc_ratio column

# < 10 is low
# between 10 and 16( both inclusive) is medium
# between 17 and 24 is high
# greater than 24 is very high

def loan_inc_ratio_category(n):
if n < 10:
return 'low'
elif n >= 10 and n < 17:
return 'medium'
elif n >= 17 and n < 25:
return 'high'
else:
return 'very high'
loan['categorised_loan_inc_ratio'] = loan['loan_inc_ratio'].apply(loan_inc_ratio_categor
loan.head()

Out[52]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65 B B2 10 RENT

1 2500 2500 2500.0 60 15.27 C C4 0 RENT

2 2400 2400 2400.0 36 15.96 C C5 10 RENT

5 5000 5000 5000.0 36 7.90 A A4 3 RENT

6 7000 7000 7000.0 60 15.96 C C5 8 RENT

In [53]: loan['int_rate'].quantile([.25,.5,.75])

0.25 8.94
Out[53]:
0.50 11.83
0.75 14.46
Name: int_rate, dtype: float64
Loading [MathJax]/extensions/Safe.js
 In [54]: #categorise int_rate column into categorised_int_rate_perc column
# < 9% is low
# between 9% and 11% ( both inclusive ) is medium
# between 12% to 13% is high
# Greater than 14% is very high

def interest_rates(n):
if n < 9:
return 'low'
elif n >= 9 and n < 12:
return 'medium'
elif n >= 12 and n < 14:
return 'high'
else:

return 'very high'

loan['categorised_int_rate_perc'] = loan['int_rate'].apply(interest_rates)
loan.head()

Out[54]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65 B B2 10 RENT

1 2500 2500 2500.0 60 15.27 C C4 0 RENT

2 2400 2400 2400.0 36 15.96 C C5 10 RENT

5 5000 5000 5000.0 36 7.90 A A4 3 RENT

6 7000 7000 7000.0 60 15.96 C C5 8 RENT

5 rows × 21 columns

In [55]: loan['emp_length'].quantile([.25,.5,.75])

0.25 2.0
Out[55]:
0.50 4.0
0.75 9.0
Name: emp_length, dtype: float64

In [56]: #categorising emp_length column into categorised_emp_length

# < 2 is entry level
# between 2 and 4 (both inclusive ) is junior level
# between 5 and 9 is middle level
# Greater than 9 is senior level

def length_of_emp(n):
if n < 2:
return 'entry level'
elif n >= 2 and n < 4:
return 'junior level'
elif n >= 4 and n < 9:
return 'middle level'
else:
return 'senior level'
loan['categorised_emp_length'] = loan['emp_length'].apply(length_of_emp)
loan.head()

Loading [MathJax]/extensions/Safe.js
 Out[56]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65 B B2 10 RENT

1 2500 2500 2500.0 60 15.27 C C4 0 RENT

2 2400 2400 2400.0 36 15.96 C C5 10 RENT

5 5000 5000 5000.0 36 7.90 A A4 3 RENT

6 7000 7000 7000.0 60 15.96 C C5 8 RENT

5 rows × 22 columns

In [57]: loan['annual_inc'].quantile([.25, .5, .75])

0.25 42000.0
Out[57]:
0.50 60000.0
0.75 82000.0
Name: annual_inc, dtype: float64

In [58]: #categorising annual_inc column into categorised_annual_inc

# < 41000 is low inc
# between 41000 and 59000 (both inclusive ) is medium inc
# between 60000 and 82000 is high inc
# Greater than 820000 is very high inc

def annual_income(n):
if n < 41000:
return 'low income'
elif n >= 41000 and n < 60000:
return 'medium income'
elif n >= 60000 and n < 83000:
return 'high income'
else:
return 'very high income'
loan['categorised_annual_inc'] = loan['annual_inc'].apply(annual_income)
loan.head()

Out[58]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65 B B2 10 RENT

1 2500 2500 2500.0 60 15.27 C C4 0 RENT

2 2400 2400 2400.0 36 15.96 C C5 10 RENT

5 5000 5000 5000.0 36 7.90 A A4 3 RENT

6 7000 7000 7000.0 60 15.96 C C5 8 RENT

5 rows × 23 columns

In [59]: loan.categorised_annual_inc.isnull().sum()

0
Out[59]:

In [60]: loan['dti'].quantile([.25, .5, .75])

0.25 8.4075
Out[60]:
0.50 13.5400
0.75 18.6500
Name: dti, dtype: float64

Loading [MathJax]/extensions/Safe.js
 In [120… loan['loan_amnt'].quantile([.25, .5, .75])

0.25 6000.0
Out[120]:
0.50 10000.0
0.75 15000.0
Name: loan_amnt, dtype: float64

In [63]: #categorising loan_amnt into categorised loan_amnt

# < 5400 is low'
# between 5400 and 9600 (both inclusive) is medium
# between 9600 and 15000 is high
# > 15000 is very high

def loan_ammount(n):
if n < 5400:
return 'low'
elif n >= 5400 and n < 9600:
return 'medium'
elif n >= 9600 and n < 15000:
return 'high'
else:
return 'very high'
loan['categorised_loan_amnt'] = loan['loan_amnt'].apply(loan_ammount)
loan.head()

Out[63]: loan_amnt funded_amnt funded_amnt_inv term int_rate grade sub_grade emp_length home_ownership

0 5000 5000 4975.0 36 10.65 B B2 10 RENT

1 2500 2500 2500.0 60 15.27 C C4 0 RENT

2 2400 2400 2400.0 36 15.96 C C5 10 RENT

5 5000 5000 5000.0 36 7.90 A A4 3 RENT

6 7000 7000 7000.0 60 15.96 C C5 8 RENT

5 rows × 24 columns

In [64]: loan.loan_status.describe()

count 33492
Out[64]:
unique 2
top Fully Paid
freq 28724
Name: loan_status, dtype: object

In [65]: loan.columns

Index(['loan_amnt', 'funded_amnt', 'funded_amnt_inv', 'term', 'int_rate',

Out[65]:
'grade', 'sub_grade', 'emp_length', 'home_ownership', 'annual_inc',
'verification_status', 'issue_d', 'loan_status', 'purpose',
'addr_state', 'dti', 'issue_month', 'issue_year', 'loan_inc_ratio',
'categorised_loan_inc_ratio', 'categorised_int_rate_perc',
'categorised_emp_length', 'categorised_annual_inc',
'categorised_loan_amnt'],
dtype='object')

In [66]: loan.loan_status.value_counts(normalize =True)*100

Fully Paid 85.763764

Out[66]:
Charged Off 14.236236
Name: loan_status, dtype: float64

In [67]: plot = sns.catplot(data = loan, x = 'loan_status', kind = 'count', palette ='Set1', aspe
Loading [MathJax]/extensions/Safe.js
plt.title('Loan Status', fontsize = 14)
 plt.xlabel('loan_status', fontsize = 12)
plt.ylabel('count', fontsize = 12)

#print the counts

ax = plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate('{:1.1f}%'.format((p.get_height()*100/len(loan))), ((p.get_x()+ p.get_w
color = 'red', ha = 'center', va = 'bottom')
plt.show()

The above graph shows that in the data provided there are 14.2% applicants who have defaulted/charged
off

In [68]: sns.distplot(loan['loan_amnt'], bins = 6)

plt.title('Loan Amount')
plt.show()

Loading [MathJax]/extensions/Safe.js
 The graph shows that the loan amount is majorly spread around 6000 t0 16000 approx.

In [69]: sns.violinplot(loan['int_rate'], inner = 'quartile', orient='h')

plt.title('Interest Rate')
plt.show()

The above graph shows that the loan percentage is spread around 8% to 14% approx.
Loading [MathJax]/extensions/Safe.js
 In [70]: sns.distplot(loan['annual_inc'], bins = 6)
plt.title('Annual Income')
plt.show()

The above figure shows that the majority of applicants have annual income is spread around 40000USD to
90000USD

In [71]: loan.columns

Index(['loan_amnt', 'funded_amnt', 'funded_amnt_inv', 'term', 'int_rate',

Out[71]:
'grade', 'sub_grade', 'emp_length', 'home_ownership', 'annual_inc',
'verification_status', 'issue_d', 'loan_status', 'purpose',
'addr_state', 'dti', 'issue_month', 'issue_year', 'loan_inc_ratio',
'categorised_loan_inc_ratio', 'categorised_int_rate_perc',
'categorised_emp_length', 'categorised_annual_inc',
'categorised_loan_amnt'],
dtype='object')

In [72]: plot = sns.catplot(data = loan, x = 'term', kind = 'count', palette ='Set1', aspect = .5
plt.title('Loan Duration', fontsize = 14)
plt.xlabel('Term', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

#print the counts

ax = plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate('{:1.1f}%'.format((p.get_height()*100/len(loan))), ((p.get_x()+ p.get_w
color = 'red', ha = 'center', va = 'bottom')
plt.show()

Loading [MathJax]/extensions/Safe.js
 The above graph shows that 74.75 applicants have taken loan of 36 months duration

In [73]: plot = sns.catplot(data = loan , x = 'purpose',kind = 'count', palette = 'Set1', aspect

plt.title('Purpose of Loan', fontsize = 16)
plt.xlabel('purpose', fontsize = 14)
plt.ylabel('count', fontsize = 14)

ax = plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate('{:1.1f}%'.format((p.get_height()*100/len(loan))), ((p.get_x()+ p.get_w
color = 'red', ha = 'center', va = 'bottom')

The following are the loan purposes for whcih more than 5% applicants have taken the loan Debt
Consilidation 52.4% Credit card 14.2% Home Improvement 8.1% Major purchase 6.2% Small Business
5.0%

Loading [MathJax]/extensions/Safe.js
 In [74]: order_grade = ['A','B','C','D','E','F','G']
plot = sns.catplot(data = loan, x ='grade', kind = 'count', palette = 'Set1', aspect = 1
plt.title('Grade', fontsize = 14)
plt.xlabel('Grade', fontsize =12)
plt.ylabel('Count', fontsize = 12)
ax = plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate('{:1.1f}%'.format((p.get_height()*100/len(loan))), ((p.get_x()+ p.get_w
color = 'red', ha = 'center', va = 'bottom')

The above graph shows that most of the applicants fall under category B(30.3%) followed by A and C
25.6%, 20.3% respectively.

In [75]: order_year = ['2007', '2008', '2009', '2010', '2011']

plot = sns.catplot(data = loan , x = 'issue_year', kind = 'count', palette = 'Set1', asp
plt.title('Loan Issue Year', fontsize = 14)
plt.xlabel('Issue Year', fontsize =12)
plt.ylabel('Count', fontsize =12)

ax =plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x() + p.get_w
color = 'red', ha = 'center', va = 'bottom')

Loading [MathJax]/extensions/Safe.js
 As per the above graph, applicants for loan increased as the year increased, In 2011 the number of
applicants for loan was 53.7%, Since the variable issue year does not provide us any direction in the
analysis so we will use it for further analysis.

In [76]: loan.categorised_emp_length.value_counts()

middle level 10680

Out[76]:
senior level 8644
junior level 7410
entry level 6758
Name: categorised_emp_length, dtype: int64

Segmented Univariate analysis

In [96]: ordered = ['entry level', 'junior level', 'middle level', 'senior level']
plot = sns.catplot(data = loan , x = 'categorised_emp_length', kind = 'count', palette =
plt.title('Employment Length', fontsize = 14)
plt.xlabel('Employment Length', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = plot.facet_axis(0,0)

for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x() + p.get_w
color = 'red', ha = 'center', va = 'bottom')

plt.show()
Loading [MathJax]/extensions/Safe.js
 There are more number of loan applicants belonging to the middle level category(31.9%) 4 to 8 Years

In [97]: ordered = ['low','medium','high', 'very high']

plot = sns.catplot(data = loan, x = 'categorised_loan_inc_ratio', kind = 'count', palett
plt.title('Loan Income Ratio', fontsize = 14)
plt.xlabel('Loan Income Ratio', fontsize = 12)
plt.ylabel('Count', fontsize =12)

ax = plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x() + p.get_w
color = 'red', ha = 'center', va = 'bottom')

Loading [MathJax]/extensions/Safe.js
 Bivariate analysis
In [98]: loan_correlation = loan.corr()
plt.figure(figsize =(14,7))
sns.heatmap(loan_correlation, annot = True, cmap = 'RdYlGn')

<Axes: >
Out[98]:

Loading [MathJax]/extensions/Safe.js
 The above heatmap shows that the loan amount, funded amount and funded amount investment are very
closely correlated, Hence we can take any of them for our analysis.

In [99]: #Loan duration vs loan status

plot = sns.catplot(data = loan , x = 'term', hue = 'loan_status', kind = 'count', palett
plt.title('Loan Duration vs Loan Status', fontsize = 14)
plt.xlabel('Term', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x()+ p.get_wi
color = 'red', ha = 'center', va = 'bottom')

Loading [MathJax]/extensions/Safe.js
 75% of applicants opted for 36 month loan duration and 66.8% have fully paid while 7.9% are charged off.
On the other hand approx 25% applicants had 60months tenure and 6.3% were charged off.

In [100… #Verification status vs Loan status

plot = sns.catplot(data = loan , x = 'verification_status', kind = 'count', hue = 'loan_
plt.title('Verfication Status vs Loan Status', fontsize = 14)
plt.xlabel('Verification Status', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x() + p.get_w
color = 'red', ha = 'center', va = 'bottom')

Loading [MathJax]/extensions/Safe.js
 The above graph shows that the applicants whose income is verified seem to default more, hence we can
ignore this as a cause for default for the further analysis.

In [101… #Home_ownership vs Loan_status

plot = sns.catplot(data = loan, x = 'home_ownership', kind = 'count', hue = 'loan_status
plt.title('Home Ownership vs Loan Status', fontsize = 14)
plt.xlabel('Home Ownership', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x() + p.get_w
color ='red', ha= 'center', va = 'bottom')

The above graph shows the applicants who are on rent and mortgage are likely to default as compaired to
applicants who own homes.
Loading [MathJax]/extensions/Safe.js
 In [102… #Loan amount vs Loan Status
ordered = ['low', 'medium', 'high', 'very high']
g = sns.catplot(data = loan, x = 'categorised_loan_amnt', kind = 'count', hue = 'loan_st
order = ordered)
plt.title('Loan Amount vs Loan Status', fontsize = 14)
plt.xlabel('Loan Amount', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = g.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x() + p.get_w
color = 'red', ha = 'center', va = 'bottom')
plt.show()

The above graph shows the applicants having high loan amount are likely to default.

In [103… #Interest rate vs Loan status

ordered = ['low', 'medium', 'high', 'very high']
g = sns.catplot(data = loan, x = 'categorised_int_rate_perc', kind = 'count', hue = 'loa
order = ordered)
plt.title('Interest Rate vs Loan Status', fontsize = 14)
plt.xlabel('Interest Rate', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = g.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x() + p.get_w
color = 'red', ha = 'center', va = 'bottom')
plt.show()

Loading [MathJax]/extensions/Safe.js
 The above graph shows the applicants having higher interest rates are likely to default.

In [104… #Annual Income vs Loan Status

ordered = ['low income','medium income', 'high income', 'very high income']
g= sns.catplot(data = loan, x = 'categorised_annual_inc', kind = 'count', hue = 'loan_st
order = ordered)
plt.title('Annual Income vs Loan Status', fontsize = 14)
plt.xlabel('Annual Income', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = g.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x() + p.get_w
color = 'red', ha = 'center', va = 'bottom')

Loading [MathJax]/extensions/Safe.js
 The above graph shows the applicants having higher annual income are less likely to default.

In [105… #Grade vs Loan Status

ordered = ['A', 'B', 'C', 'D', 'E', 'F', 'G']
n = sns.catplot(data = loan , x = 'grade', kind = 'count', hue = 'loan_status', palette
plt.title('Grade vs Loan Status', fontsize = 14)
plt.xlabel('Grade', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = n.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x()+p.get_wid
color = 'red', ha = 'center', va = 'bottom')

In [106… #Purpose vs Loan Status

plot = sns.catplot(data = loan , x = 'purpose', kind = 'count', hue = 'loan_status', pal
Loading [MathJax]/extensions/Safe.js
 plt.title('Purpose vs Loan Status', fontsize = 14)
plt.xlabel('Purpose', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = plot.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x() + p.get_w
color = 'red', ha = 'center', va= 'bottom')

plt.show()

In [107… #Loan to income vs Loan Status

ordered = ['low', 'medium', 'high', 'very high']
g = sns.catplot(data = loan, x = 'categorised_loan_inc_ratio', kind = 'count', hue = 'lo
order = ordered)
plt.title('Loan To Income vs Loan Status', fontsize = 14)
plt.xlabel('Loan To Income', fontsize = 12)
plt.ylabel('Count', fontsize = 12)

ax = g.facet_axis(0,0)
for p in ax.patches:
ax.annotate("{:1.1f}%".format((p.get_height()*100/len(loan))), ((p.get_x()+ p.get_wi
color = 'red', ha = 'center', va = 'bottom')
plt.show()

The above graph

Loading [MathJax]/extensions/Safe.js shows that the applicants having very high loan to income ratio are likely to default
 Bivariate Analysis outcome
Following are the type of applicants who are likely to default:

Applicants with a loan term of 36 months(7.4%)

Applicants who stay in rented houses and followed by applicants who stay at mortgage (7.0%) and
(6.1%) respectively
Applicants who have taken very high loan amount (>= 15000) (5.1%)
Applicants paying very (> = 14%) interest rates default more (6.6%)
Applicants with low annual income (41000) (4.2%)
Applicants identified as grade B tend to default more (3.6%) followed by C , D (3.4%) and (2.9%)
respectively
Applicants with high income to debt ratio ( between 13 and 18 inclusive) (4.5%)
Applicants who have very high loan to income ratio (>=25) (5.5%)
The default rate in terms of purpose of loan is as follows: Debt Consolidation 8.0% Credit Card 1.5%
Small Business 1.4% Home Improvement 1.0% Major Purchase 0.6%

Multivariate Analysis: Analysing more than 2

columns
In [108… #Loan status vs Purpose vs Loan amount

selected_purposes = ['debt_consolidation', 'credit_card', 'small_business', 'home_improv

filtered_df = loan[loan['purpose'].isin(selected_purposes)]

sns.catplot(y = 'loan_amnt', x = 'purpose', hue ='loan_status', data = filtered_df, kind

,height = 4)
plt.show()

In [109… #Annual income vs Loan amount vs purpose for charged off

df = filtered_df[filtered_df['loan_status'] == 'Charged Off']
ordered = ['low income', 'medium income', 'high income', 'very high income']
sns.catplot(data = df, x = 'categorised_annual_inc', y = 'loan_amnt', hue = 'purpose', a
palette = 'Set1', order = ordered, kind = 'bar')
plt.title('Annual Income vs Loan Amount vs Purpose of Defaulted',fontsize = 14)
plt.xlabel('Annual income',fontsize = 12)
Loading [MathJax]/extensions/Safe.js
 plt.ylabel('Loan Amount', fontsize = 12)
plt.show()

In [110… #Employment length vs loan Amount vs Purpose for charged off loan status
ordered = ['entry level', 'junior level', 'middle level', 'senior level']
sns.catplot(data = df, x = 'categorised_emp_length', y = 'loan_amnt', hue = 'purpose', a
palette = 'Set1', kind = 'bar', order = ordered)
plt.title('Employment Length vs Loan Amount vs Purpose of Defaulted',fontsize = 14)
plt.xlabel('Employment Length',fontsize = 12)
plt.ylabel('Loan Amount', fontsize = 12)
plt.show()

In [111… #Term vs loan Amount vs Purpose for charged off loan status
sns.catplot(data = df, x = 'term', y = 'loan_amnt', hue = 'purpose', aspect = 2.5,
palette = 'Set1', kind = 'bar')
plt.title('Term vs Loan Amount vs Purpose of Defaulted',fontsize = 14)
plt.xlabel('Term',fontsize = 12)
plt.ylabel('Loan Amount', fontsize = 12)
plt.show()

Loading [MathJax]/extensions/Safe.js
 Multivariate analysis outcome:
Following are the loan applicants who are likely to default: Applicants who take loan for small business.
Applicants whose annual income is in the category of low and medium have defaulted moreon small
business, while as the applicants who are categorised in high and very have defaulted in small business
and debt consolidation. Irrespective of employment length, those who have taken loan for small business
have defaulted. Applicants with a 60 months term duration for small business.

Hence we can infer that default rate is more for loan taken on small business followed by debt
consolidation.

In [112… #Stack Bar chart and Line Chart

def change_width(ax, new_value):

for patch in ax.patches:
patch.set_width(new_value)
patch.set_x(patch.get_x() + (patch.get_width() - new_value) * 0.5)

def plot_map(crosstab, partialtitle, label_name, value):

# Extract relevant columns from crosstab
lineplot = crosstab[['percentage_defaulted']]
barPlot = crosstab[['Charged Off', 'Fully Paid']]

# Plot line and bar charts on the same axes

ax = lineplot.plot(figsize=(20, 8), marker='o', color='b')
ax3 = barPlot.plot(kind='bar', ax=ax, rot=1, secondary_y=True, stacked=True, figsize

# Set plot title and labels

ax.set_title(f'{partialtitle} vs Percentage Default', fontsize=14)
ax.set_xlabel(label_name, fontsize=14)
ax.set_ylabel('Percentage of Default', color='b', fontsize=13)
ax3.set_ylabel('Number of Applicants', color='g', fontsize=13)

# Adjust bar width using the change_width function

change_width(ax3, value)
plt.show()

# Sample code for testing

filter_states = loan['addr_state'].value_counts()
filter_states = filter_states[filter_states < 100]
filter_states_df = loan[~loan['addr_state'].isin(filter_states.index)]

filter_states_crosstab = pd.crosstab(filter_states_df['addr_state'], filter_states_df['l

filter_states_crosstab.drop(filter_states_crosstab.tail(1).index, inplace=True)
filter_states_crosstab['percentage_defaulted'] = round(100 * (filter_states_crosstab['Ch

# Display crosstab
display(filter_states_crosstab)

# Plot the map

plot_map(filter_states_crosstab, 'State', 'state', 0.50)

Loading [MathJax]/extensions/Safe.js
 loan_status Charged Off Fully Paid All percentage_defaulted

addr_state

AL 46 338 384 12.0

AR 21 187 208 10.0

AZ 99 616 715 14.0

CA 951 5103 6054 16.0

CO 80 600 680 12.0

CT 82 550 632 13.0

DC 13 184 197 7.0

FL 433 1958 2391 18.0

GA 179 1003 1182 15.0

HI 24 120 144 17.0

IL 178 1123 1301 14.0

KS 22 199 221 10.0

KY 39 246 285 14.0

LA 45 327 372 12.0

MA 133 994 1127 12.0

MD 138 754 892 15.0

MI 80 515 595 13.0

MN 72 451 523 14.0

MO 98 499 597 16.0

NC 103 547 650 16.0

NH 23 116 139 17.0

NJ 252 1323 1575 16.0

NM 29 135 164 18.0

NV 87 323 410 21.0

NY 417 2724 3141 13.0

OH 131 908 1039 13.0

OK 35 213 248 14.0

OR 66 322 388 17.0

PA 149 1109 1258 12.0

RI 24 140 164 15.0

SC 48 330 378 13.0

TX 256 2042 2298 11.0

UT 33 196 229 14.0

VA 147 1063 1210 12.0

WA 104 598 702 15.0

WI 54 341 395 14.0

WV 18 134 152 12.0

Loading [MathJax]/extensions/Safe.js
 In [113… loan.categorised_int_rate_perc.value_counts()

medium 9430
Out[113]:
very high 9366
low 8382
high 6314
Name: categorised_int_rate_perc, dtype: int64

In [114… loan.categorised_annual_inc.isnull().sum()

0
Out[114]:

In [115… # Define the order of categories

order_category = ['low', 'medium', 'high', 'very high']

# Create a copy of the DataFrame and categorize the column

int_rate_loan_df = loan.copy()
int_rate_loan_df['categorised_int_rate_perc'] = pd.Categorical(int_rate_loan_df['categor

# Sort DataFrame by the categorized column

int_rate_loan_df = int_rate_loan_df.sort_values('categorised_int_rate_perc')

# Create a cross-tabulation
int_rate_crosstab = pd.crosstab(int_rate_loan_df['categorised_int_rate_perc'], int_rate_
int_rate_crosstab.drop(int_rate_crosstab.tail(1).index, inplace=True)

# Calculate the percentage defaulted

int_rate_crosstab['percentage_defaulted'] = round(100 * (int_rate_crosstab['Charged Off'

# Display the cross-tabulation

display(int_rate_crosstab)

# Plot the map

plot_map(int_rate_crosstab, 'int Rate', 'int rate', .15)

loan_status Charged Off Fully Paid All percentage_defaulted

categorised_int_rate_perc

low 465 7917 8382 5.55

medium 1106 8324 9430 11.73

high 975 5339 6314 15.44

very high 2222 7144 9366 23.72

Loading [MathJax]/extensions/Safe.js
 In [116… # filter_df -- percentage default for purpose

#create a crosstab
purposecrosstab = pd.crosstab(filtered_df['purpose'],filtered_df['loan_status'], margins
purposecrosstab.drop(purposecrosstab.tail(1).index, inplace = True)
purposecrosstab['percentage_defaulted'] = round(100*((purposecrosstab['Charged Off']/pur
#Display crosstab
display(purposecrosstab)
#plot the map
plot_map(purposecrosstab, 'Purpose', 'purpose', .25)

loan_status Charged Off Fully Paid All percentage_defaulted

purpose

credit_card 506 4356 4862 10.41

debt_consolidation 2665 14875 17540 15.19

home_improvement 321 2386 2707 11.86

major_purchase 209 1851 2060 10.15

small_business 459 1222 1681 27.31

Applicants who have taken loan for small business (27.31) tend to default more

In [117… #Percentage default for Grade

#Sort Grade based on custom sorting

grade_df = loan
grade_df.sort_values(['grade'])
Loading [MathJax]/extensions/Safe.js
 sub_grade_df = loan
sub_grade_df.sort_values(['sub_grade'])

#create a cross tab function

grade_crosstab = pd.crosstab(grade_df['grade'], grade_df['loan_status'], margins = True)

grade_crosstab.drop(grade_crosstab.tail(1).index, inplace = True)
grade_crosstab['percentage_defaulted'] = round(100*((grade_crosstab['Charged Off']/ grad
display(grade_crosstab)

sub_grade_crosstab = pd.crosstab(sub_grade_df['sub_grade'],sub_grade_df['loan_status'],
sub_grade_crosstab.drop(sub_grade_crosstab.tail(1).index, inplace = True)
sub_grade_crosstab['percentage_defaulted'] = round(100*((sub_grade_crosstab['Charged Off
display(sub_grade_crosstab)

plot_map(grade_crosstab, 'Grade', 'grade', .10)

plot_map(sub_grade_crosstab, 'Sub Grade', 'sub grade', .20)

loan_status Charged Off Fully Paid All percentage_defaulted

grade

A 481 8101 8582 5.60

B 1194 8959 10153 11.76

C 1126 5665 6791 16.58

D 961 3500 4461 21.54

E 628 1741 2369 26.51

F 287 578 865 33.18

G 91 180 271 33.58

Loading [MathJax]/extensions/Safe.js
 loan_status Charged Off Fully Paid All percentage_defaulted

sub_grade

A1 23 919 942 2.44

A2 58 1218 1276 4.55

A3 74 1447 1521 4.87

A4 150 2323 2473 6.07

A5 176 2194 2370 7.43

B1 131 1403 1534 8.54

B2 186 1558 1744 10.67

B3 293 2175 2468 11.87

B4 280 1841 2121 13.20

B5 304 1982 2286 13.30

C1 272 1493 1765 15.41

C2 272 1421 1693 16.07

C3 230 1078 1308 17.58

C4 177 857 1034 17.12

C5 175 816 991 17.66

D1 144 661 805 17.89

D2 226 887 1113 20.31

D3 216 765 981 22.02

D4 191 631 822 23.24

D5 184 556 740 24.86

E1 171 475 646 26.47

E2 144 402 546 26.37

E3 105 360 465 22.58

E4 113 270 383 29.50

E5 95 234 329 28.88

F1 82 189 271 30.26

F2 63 149 212 29.72

F3 44 106 150 29.33

F4 49 83 132 37.12

F5 49 51 100 49.00

G1 28 59 87 32.18

G2 25 44 69 36.23

G3 18 23 41 43.90

G4 10 38 48 20.83

G5 10 16 26 38.46

Loading [MathJax]/extensions/Safe.js
 From the above graph it can be observed that probability of applicants increases with the grades from A to
G

In [118… #Percentage default for annual income

#sort categorised annual income on custom sorting

order_category = ['low income', 'medium income', 'high income', 'very high income']
annual_inc_df = loan
annual_inc_df.categorised_annual_inc = annual_inc_df.categorised_annual_inc.astype('cate
annual_inc_df.categorised_annual_inc.cat.set_categories(order_category, inplace = True)
annual_inc_df.sort_values(['categorised_annual_inc'])

annual_inc_crosstab = pd.crosstab(annual_inc_df['categorised_annual_inc'],annual_inc_df[
annual_inc_crosstab.drop(annual_inc_crosstab.tail(1).index, inplace = True)
annual_inc_crosstab['percentage_defaulted'] = round(100*((annual_inc_crosstab['Charged O
display(annual_inc_crosstab)
plot_map(annual_inc_crosstab, 'Annual income', 'annual income', .10)

loan_status Charged Off Fully Paid All percentage_defaulted

categorised_annual_inc

low income 1406 6686 8092 17.38

medium income 1247 7239 8486 14.69

high income 1208 7414 8622 14.01

very high income 907 7385 8292 10.94

Loading [MathJax]/extensions/Safe.js
 In [119… emp_length_df = loan
emp_length_df.sort_values(['emp_length'])

emp_length_crosstab = pd.crosstab(emp_length_df['emp_length'],emp_length_df['loan_status
emp_length_crosstab.drop(emp_length_crosstab.tail(1).index, inplace = True)
emp_length_crosstab['percentage_defaulted'] = round(100*((emp_length_crosstab['Charged O
display(emp_length_crosstab)

plot_map(emp_length_crosstab, 'Emp length', 'emp length', .20)

loan_status Charged Off Fully Paid All percentage_defaulted

emp_length

0 561 3408 3969 14.13

1 400 2389 2789 14.34

2 485 3321 3806 12.74

3 489 3115 3604 13.57

4 406 2600 3006 13.51

5 414 2458 2872 14.42

6 276 1696 1972 14.00

7 235 1303 1538 15.28

8 184 1108 1292 14.24

9 144 957 1101 13.08

10 1174 6369 7543 15.56

Loading [MathJax]/extensions/Safe.js
 Following are main parameters, taken into
consideration for arriving at the analysis conclusion.
Interest ratePurpose GradeTerm Emp LengthAnnual Income

As per the analysis it can inferred that applicants who are of low income and have taken high interest loan
with longer duration for small business have more profitability of defaulting.

Hence Extra care should be taken before lending them loan.

In [ ]:

In [ ]:

Loading [MathJax]/extensions/Safe.js